What Is Agentic Commerce? A Plain-English Guide

The short version

Agentic commerce is what happens when an AI assistant does the shopping for you. Instead of typing keywords into a search box and clicking through ten tabs, you tell an AI what you need, and it comes back with specific products, answers your questions, and completes the purchase. Finding, deciding, and buying all happen in one place: the conversation.

If you have ever asked ChatGPT "what should I get my mum for her birthday" or asked Perplexity for the "best running shoes for flat feet under 8,000," you have already touched the edge of it. The next step is the AI not just naming products, but letting you buy them right there.

This is happening on tools people already use every day: ChatGPT, Google's Gemini and AI Mode, Perplexity, Claude, Amazon's Rufus, WhatsApp assistants, agentic browsers, and a growing pile of custom AI agents and apps.

Why now and not five years ago

Three things had to line up, and in the last year they did.

Behaviour moved first. People got comfortable asking AI for advice, and shopping questions came along for the ride. "Best tops under 1000." "Something nice to gift her." "Which brand has the best linen." Billions of queries a day now run through ChatGPT, Google AI, and Perplexity, and a meaningful slice of them are really product searches in disguise.

The traffic followed. Adobe has reported a steep year-over-year jump in retail traffic coming from AI sources. And the people arriving this way behave differently. They show up already informed, because they did their comparison in the chat before they ever clicked.

The plumbing arrived. OpenAI shipped the Agentic Commerce Protocol (ACP) and Google shipped the Universal Commerce Protocol (UCP). In plain terms, these are the rails that let a product be discovered and bought without the shopper ever leaving the chat.

How a purchase actually happens

Once you see the mechanics, it becomes obvious why some products get recommended and others never show up.

1. The question fans out

When someone asks for "noise cancelling headphones for travel under 3,000 with good battery life," the AI does not run one search. It quietly breaks the request into a fan of smaller ones: best noise cancelling headphones for travel, wireless over-ear under 3,000, long battery life for flights, top rated ANC, and so on.

2. Those sub-questions hit your product data

Each one gets matched against structured product feeds: category, attributes, specs, reviews, price, availability. Here is the part most brands miss. The AI does not guess, and it does not give you the benefit of the doubt. If your product data does not clearly answer those sub-questions, your product simply is not in the running, no matter how good it is.

3. The AI recommends, answers, and sells

The products that win come back as cards in the chat. The shopper asks follow-ups, like "is this safe for sensitive skin" or "what colours does it come in," and the AI answers from enriched data and pooled reviews. Then checkout, the address, the delivery choice, the payment, all happens right there. No redirect.

The four levels every product has to climb

Think of it as a ladder. Most catalogs are stuck on the bottom rung.

The reason so many catalogs stall at level one is simple. They were written for human eyes and a search box, not for an agent running structured comparisons in the background. Closing that gap is the whole job of feed optimization for AI search.

What this means if you run a brand

Three new kinds of buyers are forming, and not one of them is browsing your website:

• The AI platforms themselves (ChatGPT, Gemini, Perplexity, AI Mode), answering "what should I buy" with a short list.
• AI agents shopping on someone's behalf: gifting agents, restock bots, trip planners.
• People referred by AI, who arrive having already decided and expect to buy in one step.

Winning all three comes down to the same work: structured, enriched, trustworthy product data, synced everywhere AI is listening, plus a storefront that can actually close a sale inside a conversation. A good starting point is how to sell on ChatGPT and answer engine optimization for ecommerce.

What this means if you build AI

If you are building an agent, a custom GPT, or a consumer AI app, agentic commerce is two things at once: the way your AI becomes genuinely useful, and the way it finally earns money. One commerce integration, for example a shopping MCP server, hands your agent real products with live price and stock, native checkout, and a cut of every sale it drives. We walk through the money side in how to monetize your AI agent.

Agentic commerce vs. classic ecommerce

Common questions

Is this just conversational commerce with a new name?

Not quite. Conversational commerce, the old chat-based selling with scripted bots, was the warm-up. Agentic commerce adds real autonomy: the AI can search structured catalogs, reason over attributes, and run the transaction itself.

Do people really buy inside an AI chat?

Yes, and that is exactly what ACP and UCP were built for. Buying in the conversation removes the redirect, which is where most shoppers quietly drop off.

What should a brand do first?

Find out whether AI assistants recommend you today (see AI visibility and share of voice), then fix the data gaps keeping you invisible. Or connect infrastructure like Ziffi that does both at once.